Dynamic modeling and fault feature analysis of ball-cage scraping in flywheel bearings

Abnormal ball-cage scraping in flywheel bearings typically induces elevated friction, instability, and even complete failure. However, the characteristics of such faults remain unclear due to complex mechanisms, which pose challenges for fault diagnosis. This paper establishes a dynamic model for ball-cage scraping faults in flywheel bearings to reveal fault response and dynamic characteristics, providing theoretical guidance for diagnosis. Firstly, the scraping friction process is simulated to construct the mechanism model, where ball–cage scraping is characterized by introducing scraping-induced equivalent impact force and friction. Subsequently, a nonlinear fault dynamic model capable of reflecting fault characteristics is developed, followed by experimental validation. On this basis, the influence of various factors on dynamic characteristics is systematically investigated. The results reveal that when ball-cage scraping occurs, distinct characteristic frequencies emerge in the vibration magnitude spectrum, including ball fault frequency (BFF), cage rotational frequency (CRF), and first- and second-order BFF sidebands modulated by CRF. The amplitudes of these frequencies can reflect the severity and duration of the scraping fault and are influenced by factors such as cage whirling, load, and other operational parameters. Finally, based on the revealed evolution patterns of fault characteristics, the lifetime data of flywheel bearings is analyzed to inform engineering practices for health stage identification, thereby providing robust support for health management.